Conventionally, in a digital copying machine, a scanner and other apparatuses, an original document is read by a CCD or other devices, and the image-data signal obtained in such a reading operation is subjected to image-processing operations in order to improve the image quality. For example, in such an image-processing method, various features are extracted from image information that has been read, and based upon the features, a region-discriminating process can be carried out so that a high-quality image-processing operation is carried out by using the results of this region-discriminating process.
One of the above-mentioned image-processing methods is disclosed by, for example, "Image-Processing Apparatus" in Japanese Laid-Open Patent Publication No. 188948/1992 (Tokukaihei 4-188948). In this method, the average value of density levels in all the pixels within a specified region is given as a threshold value. Then, if the difference between the threshold value and the average dispersion value of density levels of pixels is large, the above-mentioned specified region is discriminated as a character-image region. In contrast, if the difference between the threshold value and the average dispersion value is relatively small, the specified region is discriminated as a half-tone-image region. Based upon the result of the discrimination, the image-processing apparatus carries out the image-processing.
Moreover, an image-processing method similar to the above-mentioned method is used in an "Image-Processing Apparatus" disclosed by Japanese Laid-Open patent Publication No. 270561/1992 (Tokukaihei 4-270561). In this case, two large and small threshold values THb and THc are set with respect to density levels of the pixels within a specified region. Then, portions having the density level higher than THb and portions having the density level lower than THc are discriminated as character-image regions, and portions having density levels between THb and THc are discriminated as half-tone regions.
However, in the above-mentioned conventional arrangements, depending on image-data signals within the specified region, discrimination is made so as to identify the kind of region with respect to the entire specified region. The resulting problem is that erroneous discrimination tends to occur depending on how the specified region is selected.
Moreover, in the above-mentioned conventional arrangements, only the difference between the average value and the average dispersion in the density level or only the predetermined threshold value is used as the degree of feature so as to make comparisons on the magnitude of the density level of each pixel. Therefore, since it is not possible to sufficiently discriminate features of the image region, the resulting problem is degradation in discrimination precision in discriminating the various kinds of regions.